CN112317957A - Laser welding method, laser welding apparatus, and storage medium therefor - Google Patents
Laser welding method, laser welding apparatus, and storage medium therefor Download PDFInfo
- Publication number
- CN112317957A CN112317957A CN202011072145.4A CN202011072145A CN112317957A CN 112317957 A CN112317957 A CN 112317957A CN 202011072145 A CN202011072145 A CN 202011072145A CN 112317957 A CN112317957 A CN 112317957A
- Authority
- CN
- China
- Prior art keywords
- data set
- sample
- laser welding
- class
- cluster
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000003466 welding Methods 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000003860 storage Methods 0.000 title claims abstract description 12
- 238000012549 training Methods 0.000 claims abstract description 44
- 238000013508 migration Methods 0.000 claims abstract description 13
- 230000005012 migration Effects 0.000 claims abstract description 13
- 238000013136 deep learning model Methods 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 19
- 238000012544 monitoring process Methods 0.000 claims description 12
- 238000012935 Averaging Methods 0.000 claims description 6
- 230000005484 gravity Effects 0.000 claims description 6
- 230000007547 defect Effects 0.000 abstract description 5
- 238000013526 transfer learning Methods 0.000 abstract description 4
- 230000000694 effects Effects 0.000 description 11
- 230000002411 adverse Effects 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000001066 destructive effect Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K26/00—Working by laser beam, e.g. welding, cutting or boring
- B23K26/20—Bonding
- B23K26/21—Bonding by welding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Optics & Photonics (AREA)
- Plasma & Fusion (AREA)
- Mechanical Engineering (AREA)
- Laser Beam Processing (AREA)
Abstract
The invention discloses a laser welding method, a laser welding device and a storage medium thereof, wherein the method comprises the steps of collecting a bad factor sample set in the laser welding process as a target data set; confirming an existing sample set with the highest similarity to a target data set as a source data set; continuously pre-training the deep learning model by using a source data set to obtain a first pre-training model; performing structural adjustment on the first pre-training model by using the target data set to obtain a second pre-training model; and transferring the parameters of the first pre-training model to the second pre-training model to obtain a final model. The source data set which is matched with the target data set in extremely high similarity can be confirmed, and negative migration is avoided; the defect that a small amount of sample data is easy to over-fit is overcome and the convergence speed of the model is accelerated by utilizing a transfer learning mode.
Description
Technical Field
The invention relates to the field of laser welding, in particular to a laser welding method, a laser welding device and a storage medium thereof.
Background
The laser welding technology is widely and deeply applied to the fields of aerospace, automobile manufacturing, electronic consumer products and the like. However, various undesirable factors such as surface impurities and abrasion of materials, wrong process parameters, human operation error factors, etc. exist during the laser welding process, and the occurrence of the undesirable factors may cause defects to be continuously generated with a high probability. The laser welding process is accompanied with the release of a large number of sound, light, electricity and heat signals, and the monitoring of the series of signals can be realized by depending on various sensors. In the monitoring process, the correlation between the signal intensity change and various adverse factors can be established by using a deep learning technology. The deep learning technology needs a large amount of sample data to obtain a good performance effect, but in some laser welding processing environments, the acquisition of the sample data is very difficult, the acquisition of bad samples needs to be realized by resorting to manual sampling and destructive acquisition methods, and the acquisition of the bad samples always needs a large amount of manpower and cost, so that the acquisition of a large amount of sample data is difficult. Under the condition that sample data is less, the monitoring effect of the existing deep learning model on laser welding is greatly limited.
Disclosure of Invention
The present invention is directed to solving at least one of the problems of the prior art, and provides a laser welding method, apparatus and storage medium thereof.
The technical scheme adopted by the invention for solving the problems is as follows:
in a first aspect of the present invention, a laser welding method comprises the steps of:
collecting a bad factor sample set in a laser welding process as a target data set, wherein the bad factor sample set comprises a plurality of kinds of bad factor samples;
confirming an existing sample set with the highest similarity with the target data set from a plurality of existing sample sets as a source data set;
continuously pre-training a deep learning model for monitoring laser welding by using the source data set until the model loss reaches the minimum value to obtain a first pre-training model;
performing structural adjustment on the first pre-training model by using the target data set to obtain a second pre-training model;
and transferring the parameters of the first pre-training model to the second pre-training model to obtain a final model.
According to the first aspect of the present invention, the identifying, from the plurality of existing sample sets, the data set with the highest similarity to the target data set as the source data set includes the following steps:
dividing each existing sample set into a plurality of classes according to class labels, and confirming a representative sample of each class of the existing sample set;
dividing the target data set into a plurality of classes according to class labels, and confirming a representative sample of each class of the target data set;
confirming the distance between each representative sample of the existing sample set and all representative samples of the target data set, and taking the average value of p distances with the smallest median value of all distances as a distance comparison value of the existing sample set;
comparing the distance comparison values of all the existing sample sets, and taking the existing sample set corresponding to the minimum distance comparison value as the source data set;
wherein the representative samples correspond to the classes one to one.
According to a first aspect of the invention, validating a representative sample of classes comprises the steps of:
identifying a center sample at a center position of the class;
confirming a dynamic regular matrix of the central sample and other samples in the class, and obtaining a shortest path with minimum regular cost through the dynamic regular matrix, wherein elements of the dynamic regular matrix are Euclidean distances between the central sample and other samples in the class;
and carrying out gravity center averaging on each coordinate on the shortest path to obtain a new path, and taking the new path as a representative sample.
According to a first aspect of the invention, said confirming the central sample at the central position of the class comprises the steps of:
randomly selecting k elements in the class as a cluster center;
a cluster grouping step: calculating the Euclidean distance between other elements in the class and the center of each cluster, and dividing the other elements in the class into the cluster corresponding to the cluster center with the minimum Euclidean distance according to the nearest neighbor principle;
confirming a new cluster center: calculating a criterion value of each element in each cluster, and taking the element with the minimum criterion value as a new cluster center;
and repeating the cluster grouping step and the new cluster center confirming step until the elements serving as the cluster centers do not change any more, and outputting k elements serving as the cluster centers as the center samples.
In a second aspect of the present invention, a laser welding apparatus includes:
the laser welding system comprises a sample acquisition module, a data acquisition module and a data acquisition module, wherein the sample acquisition module is used for collecting a bad factor sample set in a laser welding process as a target data set, and the bad factor sample set comprises multiple kinds of bad factor samples;
the source data set acquisition module is used for confirming an existing sample set with the highest similarity to the target data set from the existing sample sets as a source data set;
the pre-training module is used for continuously pre-training the deep learning model for monitoring laser welding by using the source data set until the model loss reaches the minimum value, so as to obtain a first pre-training model;
the adjusting module is used for carrying out structural adjustment on the first pre-training model by using the target data set to obtain a second pre-training model;
and the parameter migration module is used for migrating the parameters of the first pre-training model to the second pre-training model to obtain a final model.
According to a second aspect of the present invention, the source data set acquisition module performs the following steps:
dividing each existing sample set into a plurality of classes according to class labels, and confirming a representative sample of each class of the existing sample set;
dividing the target data set into a plurality of classes according to class labels, and confirming a representative sample of each class of the target data set;
confirming the distance between each representative sample of the existing sample set and all representative samples of the target data set, and taking the average value of p distances with the smallest median value of all distances as a distance comparison value of the existing sample set;
comparing the distance comparison values of all the existing sample sets, and taking the existing sample set corresponding to the minimum distance comparison value as the source data set;
wherein the representative samples correspond to the classes one to one.
According to a second aspect of the present invention, the source data set acquisition module further performs the following steps to identify a representative sample of classes:
identifying a center sample at a center position of the class;
confirming a dynamic regular matrix of the central sample and other samples in the class, and obtaining a shortest path with minimum regular cost through the dynamic regular matrix, wherein elements of the dynamic regular matrix are Euclidean distances between the central sample and other samples in the class;
and carrying out gravity center averaging on each coordinate on the shortest path to obtain a new path, and taking the new path as a representative sample.
According to a second aspect of the present invention, the source data set acquisition module further performs the following steps to identify a center sample at the center position of the class:
randomly selecting k elements in the class as a cluster center;
a cluster grouping step: calculating the Euclidean distance between other elements in the class and the center of each cluster, and dividing the other elements in the class into the cluster corresponding to the cluster center with the minimum Euclidean distance according to the nearest neighbor principle;
confirming a new cluster center: calculating a criterion value of each element in each cluster, and taking the element with the minimum criterion value as a new cluster center;
and repeating the cluster grouping step and the new cluster center confirming step until the elements serving as the cluster centers do not change any more, and outputting k elements serving as the cluster centers as the center samples.
In a third aspect of the present invention, a laser welding apparatus includes a plurality of sensors, a processor, and a memory for storing executable instructions executable on the processor, the plurality of sensors and the memory each coupled to the processor; wherein the processor executes the executable instructions to perform the laser welding method according to the first aspect of the invention; wherein the sensor is used for acquiring signals in the laser welding process.
In a fourth aspect of the present invention, a storage medium stores executable instructions that are executable by a computer to cause the computer to perform the laser welding method according to the first aspect of the present invention.
The scheme at least has the following beneficial effects: the source data set which is matched with the target data set in extremely high similarity can be confirmed, negative migration is avoided, and the parameter migration effect is improved, so that the final model convergence effect is better; the defect that a small amount of sample data is easy to over-fit is overcome and the convergence speed of the model is accelerated by utilizing a transfer learning mode.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The invention is further illustrated with reference to the following figures and examples.
FIG. 1 is a flow chart of a laser welding method according to an embodiment of the present invention;
FIG. 2 is a detailed flowchart of step S200 in FIG. 1;
fig. 3 is a structural view of a laser welding apparatus according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to the present preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to like elements throughout.
In the description of the present invention, it should be understood that the orientation or positional relationship referred to in the description of the orientation, such as the upper, lower, front, rear, left, right, etc., is based on the orientation or positional relationship shown in the drawings, and is only for convenience of description and simplification of description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention.
In the description of the present invention, the meaning of a plurality of means is one or more, the meaning of a plurality of means is two or more, and larger, smaller, larger, etc. are understood as excluding the number, and larger, smaller, inner, etc. are understood as including the number. If the first and second are described for the purpose of distinguishing technical features, they are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated or implicitly indicating the precedence of the technical features indicated.
In the description of the present invention, unless otherwise explicitly limited, terms such as arrangement, installation, connection and the like should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above terms in the present invention in combination with the specific contents of the technical solutions.
Referring to fig. 1, a first embodiment of the present invention provides a laser welding method. The laser welding method comprises the following steps:
s100, collecting a bad factor sample set in a laser welding process as a target data set, wherein the bad factor sample set comprises multiple kinds of bad factor samples;
s200, confirming an existing sample set with the highest similarity to a target data set from a plurality of existing sample sets as a source data set;
s300, continuously pre-training a deep learning model for monitoring laser welding by using a source data set until the model loss reaches the minimum value to obtain a first pre-training model;
s400, performing structural adjustment on the first pre-training model by using a target data set to obtain a second pre-training model;
and S500, transferring the parameters of the first pre-training model to the second pre-training model to obtain a final model.
In the embodiment, the laser welding method can confirm the source data set which is matched with the target data set in a very high similarity, avoid negative migration, improve the parameter migration effect and enable the final model convergence effect to be better; the defect that a small amount of sample data is easy to over-fit is overcome by using a transfer learning mode, and the convergence speed of the model is increased; the relevance between the signal intensity change established by the final model and various adverse factors is higher, and the monitoring on the adverse factors of laser welding is facilitated.
The existing sample set is a data sample in a large database, and may be an adverse factor sample collected in the past. The samples in the large database are partially related to the adverse factor samples and partially unrelated to the adverse factor samples.
It should be noted that the deep learning model for monitoring laser welding is not limited herein. In this embodiment, a deep learning model of the CNN framework is employed. Of course, in other embodiments, other deep learning models may be used.
Referring to fig. 2, further, in step S200, identifying an existing sample set with the highest similarity to the target data set as the source data set from the existing sample sets includes the following steps:
step S210, dividing each existing sample set into a plurality of classes according to class labels, and confirming a representative sample of each class of the existing sample set; wherein, the representative samples correspond to the classes one by one;
step S220, dividing the target data set into a plurality of classes according to class labels, and confirming a representative sample of each class of the second data set; wherein, the representative samples correspond to the classes one by one;
it should be noted that, the data samples of the existing sample set and the target data set are all labeled with class labels;
step S230, confirming the distance between each representative sample of the existing sample set and all representative samples of the target data set, and taking the average value of p distances with the minimum median value of all distances as a distance comparison value of the existing sample set; the distance is specifically European distance;
in step S230, if an existing sample set has d classes, i.e. d representative samples, the target data set has e representative samples; constructing d-e Euclidean distances for an existing sample set, and taking the average value of the p distances with the minimum value from the d-e Euclidean distances as a distance comparison value of the existing sample set; repeating the step S230 to obtain the distance contrast values of all the existing sample sets;
step S240, comparing the distance comparison values of all the existing sample sets, and taking the existing sample set corresponding to the smallest distance comparison value as the source data set. The existing sample set corresponding to the minimum distance contrast value is the data set with the highest similarity with the target data set.
It should be noted that if the similarity between the source data set and the target data set is not sufficient, a negative migration situation is likely to occur. The source data set and the target data set for parameter migration are different but have certain correlation, and the higher the similarity between the source data set and the target data set, the better the parameter migration effect. The data set with the highest similarity to the target data set can be accurately found through the similarity measurement method.
Further, in step S210 and step S220, confirming the representative sample of the class includes the steps of:
confirming a central sample at the central position of the class;
confirming a dynamic regular matrix of the central sample and other samples in the class, and obtaining a shortest path with the minimum regular cost through the dynamic regular matrix, wherein elements of the dynamic regular matrix are Euclidean distances between the central sample and other samples in the class;
and averaging the centers of gravity of each coordinate on the shortest path to obtain a new path, and taking the new path as a representative sample.
It should be noted that, for a dynamic regularization matrix, there are multiple paths from the origin to a specific point, and one of the paths is defined as W ═ W1，w2，…，wK，max(m,n)≤K≤m+n-1；w1Representing a first point through which the path passes; the regular cost is
Where Q is the center sample and C is the other samples in the class. Center of gravityAverage is expressed as bartcenter { w }1，…，wK}＝(w1+…+wK)/K。
Further, confirming the center sample at the center position of the class comprises the steps of:
randomly selecting k elements in the class as a cluster center;
a cluster grouping step: calculating the Euclidean distance between other elements in the class and the center of each cluster, and dividing the other elements in the class into the cluster corresponding to the cluster center with the minimum Euclidean distance to the element according to the nearest neighbor principle; i.e. all elements in a class are divided into k clusters;
confirming a new cluster center: calculating a criterion value of each element in each cluster, and taking the element with the minimum criterion value as a new cluster center;
and repeating the cluster grouping step and the step of confirming the center of a new cluster until the elements as the centers of the clusters are not changed any more, and outputting k elements as the centers of the clusters as center samples.
Referring to fig. 3, a second embodiment of the present invention provides a laser welding apparatus. The laser welding device includes:
the sample acquisition module 10 is used for collecting a bad factor sample set in the laser welding process as a target data set, wherein the bad factor sample set comprises multiple kinds of bad factor samples;
a source data set obtaining module 20, configured to determine, from the existing sample sets, an existing sample set with the highest similarity to the target data set as a source data set;
the pre-training module 30 is configured to continuously pre-train the deep learning model for monitoring laser welding by using the source data set until the model loss reaches a minimum value, so as to obtain a first pre-training model;
an adjusting module 40, configured to perform structural adjustment on the first pre-training model using the target data set to obtain a second pre-training model;
and the parameter migration module 50 is configured to migrate the parameters of the first pre-training model to the second pre-training model to obtain a final model.
In this embodiment, the laser welding apparatus, using the laser welding method as in the method embodiment, can confirm the source data set that matches with the target data set with a very high similarity, avoid negative migration, improve the parameter migration effect, and make the final model convergence effect better; the defect that a small amount of sample data is easy to over-fit is overcome by using a transfer learning mode, and the convergence speed of the model is increased; the relevance between the signal intensity change established by the final model and various adverse factors is higher, and the monitoring on the adverse factors of laser welding is facilitated.
Further, the source data set acquiring module 20 performs the following steps:
dividing each existing sample set into a plurality of classes according to class labels, and confirming a representative sample of each class of the existing sample set;
dividing the target data set into a plurality of classes according to class labels, and confirming a representative sample of each class of the second data set;
confirming the distance between each representative sample of the existing sample set and all representative samples of the target data set, and taking the average value of p distances with the minimum median value of all distances as a distance comparison value of the existing sample set;
comparing the distance comparison values of all the existing sample sets, and taking the existing sample set corresponding to the minimum distance comparison value as a source data set;
wherein, the representative samples correspond to the classes one by one.
Further, the source data set acquisition module 20 performs the following steps to confirm the representative sample of the class:
confirming a central sample at the central position of the class;
confirming a dynamic regular matrix of the central sample and other samples in the class, and obtaining a shortest path with the minimum regular cost through the dynamic regular matrix, wherein elements of the dynamic regular matrix are Euclidean distances between the central sample and other samples in the class;
and averaging the centers of gravity of each coordinate on the shortest path to obtain a new path, and taking the new path as a representative sample.
Further, the source data set acquisition module 20 performs the following steps to confirm the center sample at the center position of the class:
randomly selecting k elements in the class as a cluster center;
a cluster grouping step: calculating the Euclidean distance between other elements in the class and the center of each cluster, and dividing the other elements in the class into the cluster corresponding to the cluster center with the minimum Euclidean distance according to the nearest neighbor principle;
confirming a new cluster center: calculating a criterion value of each element in each cluster, and taking the element with the minimum criterion value as a new cluster center;
and repeating the cluster grouping step and the step of confirming the center of a new cluster until the elements as the centers of the clusters are not changed any more, and outputting k elements as the centers of the clusters as center samples.
It should be noted that the laser welding apparatus, applying the laser welding method as described in the method embodiment, can perform each step of the laser welding method through cooperation of each module, and has the same technical effect, and will not be described in detail herein.
In a third aspect of the present invention, a laser welding apparatus is provided. The laser welding device comprises a plurality of sensors, a processor and a memory for storing executable instructions capable of running on the processor, wherein the plurality of sensors and the memory are connected with the processor; wherein the processor executes executable instructions to perform a laser welding method as described in method embodiments of the present invention; wherein the sensor is used for acquiring signals in the laser welding process.
The sensor comprises an acoustic sensor, a spectrometer, a photodiode, a visual sensor, a temperature sensor and the like, is used for acquiring acoustic, optical, electrical and thermal signals and signal intensity changes in the laser welding process, and can extract adverse factor samples in the signal intensity changes.
In a fourth aspect of the present invention, a storage medium stores executable instructions that are executable by a computer to cause the computer to perform a laser welding method as described in method embodiments of the present invention.
Examples of storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device.
The above description is only a preferred embodiment of the present invention, and the present invention is not limited to the above embodiment, and the present invention shall fall within the protection scope of the present invention as long as the technical effects of the present invention are achieved by the same means.
Claims (10)
1. A laser welding method, characterized by comprising the steps of:
collecting a bad factor sample set in a laser welding process as a target data set, wherein the bad factor sample set comprises a plurality of kinds of bad factor samples;
confirming an existing sample set with the highest similarity with the target data set from a plurality of existing sample sets as a source data set;
continuously pre-training a deep learning model for monitoring laser welding by using the source data set until the model loss reaches the minimum value to obtain a first pre-training model;
performing structural adjustment on the first pre-training model by using the target data set to obtain a second pre-training model;
and transferring the parameters of the first pre-training model to the second pre-training model to obtain a final model.
2. The laser welding method according to claim 1, wherein the identifying, from the plurality of existing sample sets, an existing sample set having the highest similarity to the target data set as the source data set comprises the steps of:
dividing each existing sample set into a plurality of classes according to class labels, and confirming a representative sample of each class of the existing sample set;
dividing the target data set into a plurality of classes according to class labels, and confirming a representative sample of each class of the target data set;
confirming the distance between each representative sample of the existing sample set and all representative samples of the target data set, and taking the average value of p distances with the smallest median value of all distances as a distance comparison value of the existing sample set;
comparing the distance comparison values of all the existing sample sets, and taking the existing sample set corresponding to the minimum distance comparison value as the source data set;
wherein the representative samples correspond to the classes one to one.
3. The laser welding method according to claim 2, wherein identifying representative samples of classes comprises the steps of:
identifying a center sample at a center position of the class;
confirming a dynamic regular matrix of the central sample and other samples in the class, and obtaining a shortest path with minimum regular cost through the dynamic regular matrix, wherein elements of the dynamic regular matrix are Euclidean distances between the central sample and other samples in the class;
and carrying out gravity center averaging on each coordinate on the shortest path to obtain a new path, and taking the new path as a representative sample.
4. The laser welding method according to claim 3, wherein the confirming of the center sample at the center position of the class comprises the steps of:
randomly selecting k elements in the class as a cluster center;
a cluster grouping step: calculating the Euclidean distance between other elements in the class and the center of each cluster, and dividing the other elements in the class into the cluster corresponding to the cluster center with the minimum Euclidean distance according to the nearest neighbor principle;
confirming a new cluster center: calculating a criterion value of each element in each cluster, and taking the element with the minimum criterion value as a new cluster center;
and repeating the cluster grouping step and the new cluster center confirming step until the elements serving as the cluster centers do not change any more, and outputting k elements serving as the cluster centers as the center samples.
5. A laser welding apparatus, comprising:
the laser welding system comprises a sample acquisition module, a data acquisition module and a data acquisition module, wherein the sample acquisition module is used for collecting a bad factor sample set in a laser welding process as a target data set, and the bad factor sample set comprises multiple kinds of bad factor samples; a source data set acquisition module, configured to determine, from a plurality of existing sample sets, an existing sample set with the highest similarity to the target data set as a source data set;
the pre-training module is used for continuously pre-training the deep learning model for monitoring laser welding by using the source data set until the model loss reaches the minimum value, so as to obtain a first pre-training model;
the adjusting module is used for carrying out structural adjustment on the first pre-training model by using the target data set to obtain a second pre-training model;
and the parameter migration module is used for migrating the parameters of the first pre-training model to the second pre-training model to obtain a final model.
6. The laser welding apparatus according to claim 5, wherein the source data set acquisition module performs the steps of:
dividing each existing sample set into a plurality of classes according to class labels, and confirming a representative sample of each class of the existing sample set;
dividing the target data set into a plurality of classes according to class labels, and confirming a representative sample of each class of the target data set;
confirming the distance between each representative sample of the existing sample set and all representative samples of the target data set, and taking the average value of p distances with the smallest median value of all distances as a distance comparison value of the existing sample set;
comparing the distance comparison values of all the existing sample sets, and taking the existing sample set corresponding to the minimum distance comparison value as the source data set;
wherein the representative samples correspond to the classes one to one.
7. The laser welding apparatus of claim 6, wherein the source data set acquisition module further performs the following steps to identify a representative sample of classes:
identifying a center sample at a center position of the class;
confirming a dynamic regular matrix of the central sample and other samples in the class, and obtaining a shortest path with minimum regular cost through the dynamic regular matrix, wherein elements of the dynamic regular matrix are Euclidean distances between the central sample and other samples in the class;
and carrying out gravity center averaging on each coordinate on the shortest path to obtain a new path, and taking the new path as a representative sample.
8. The laser welding apparatus of claim 7, wherein the source data set acquisition module further performs the following steps to identify a center sample at a center position of a class: randomly selecting k elements in the class as a cluster center;
a cluster grouping step: calculating the Euclidean distance between other elements in the class and the center of each cluster, and dividing the other elements in the class into the cluster corresponding to the cluster center with the minimum Euclidean distance according to the nearest neighbor principle;
confirming a new cluster center: calculating a criterion value of each element in each cluster, and taking the element with the minimum criterion value as a new cluster center;
and repeating the cluster grouping step and the new cluster center confirming step until the elements serving as the cluster centers do not change any more, and outputting k elements serving as the cluster centers as the center samples.
9. A laser welding apparatus comprising a plurality of sensors, a processor, and a memory for storing executable instructions executable on the processor, the plurality of sensors and the memory each coupled to the processor; wherein the processor executes the executable instructions to perform the laser welding method of any one of claims 1 to 4; wherein the sensor is used for acquiring signals in the laser welding process.
10. Storage medium, characterized in that it stores executable instructions that can be executed by a computer, causing the computer to perform the laser welding method according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011072145.4A CN112317957B (en) | 2020-10-09 | 2020-10-09 | Laser welding method, laser welding apparatus, and storage medium therefor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011072145.4A CN112317957B (en) | 2020-10-09 | 2020-10-09 | Laser welding method, laser welding apparatus, and storage medium therefor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112317957A true CN112317957A (en) | 2021-02-05 |
| CN112317957B CN112317957B (en) | 2022-11-08 |
Family
ID=74313399
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011072145.4A Active CN112317957B (en) | 2020-10-09 | 2020-10-09 | Laser welding method, laser welding apparatus, and storage medium therefor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112317957B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108846444A (en) * | 2018-06-23 | 2018-11-20 | 重庆大学 | The multistage depth migration learning method excavated towards multi-source data |
| CN109271522A (en) * | 2018-11-20 | 2019-01-25 | 深圳大学 | Comment sensibility classification method and system based on depth mixed model transfer learning |
| CN109754017A (en) * | 2019-01-09 | 2019-05-14 | 西北工业大学 | Based on separable three-dimensional residual error network and transfer learning hyperspectral image classification method |
| CN110555467A (en) * | 2019-08-13 | 2019-12-10 | 深圳创新奇智科技有限公司 | industrial data classification method based on model migration |
| CN110728294A (en) * | 2019-08-30 | 2020-01-24 | 北京影谱科技股份有限公司 | Cross-domain image classification model construction method and device based on transfer learning |
| WO2020077693A1 (en) * | 2018-10-17 | 2020-04-23 | 深圳市微埃智能科技有限公司 | Transfer learning-based method and device for recommending welding process parameters, and robot |
-
2020
- 2020-10-09 CN CN202011072145.4A patent/CN112317957B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108846444A (en) * | 2018-06-23 | 2018-11-20 | 重庆大学 | The multistage depth migration learning method excavated towards multi-source data |
| WO2020077693A1 (en) * | 2018-10-17 | 2020-04-23 | 深圳市微埃智能科技有限公司 | Transfer learning-based method and device for recommending welding process parameters, and robot |
| CN109271522A (en) * | 2018-11-20 | 2019-01-25 | 深圳大学 | Comment sensibility classification method and system based on depth mixed model transfer learning |
| CN109754017A (en) * | 2019-01-09 | 2019-05-14 | 西北工业大学 | Based on separable three-dimensional residual error network and transfer learning hyperspectral image classification method |
| CN110555467A (en) * | 2019-08-13 | 2019-12-10 | 深圳创新奇智科技有限公司 | industrial data classification method based on model migration |
| CN110728294A (en) * | 2019-08-30 | 2020-01-24 | 北京影谱科技股份有限公司 | Cross-domain image classification model construction method and device based on transfer learning |
Non-Patent Citations (2)
| Title |
|---|
| 唐亚，白治江: "基于LDA主题相似度的SVM迁移学习", 《微型机与应用》 * |
| 黄旭丰: "基于深度迁移学习的焊接质量在线监测方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112317957B (en) | 2022-11-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105608471B (en) | Robust direct-push label estimation and data classification method and system | |
| JP2020126601A (en) | Fault detecting method and fault detecting device | |
| CN113807350A (en) | Target detection method, device, equipment and storage medium | |
| CN113591948A (en) | Defect pattern recognition method and device, electronic equipment and storage medium | |
| CN117670820B (en) | Plastic film production defect detection method and system | |
| CN114548199A (en) | Multi-sensor data fusion method based on deep migration network | |
| CN110634198B (en) | Industrial system layered fault diagnosis method based on regular polycell filtering | |
| CN116823793A (en) | Device defect detection method, device, electronic device and readable storage medium | |
| KR101782364B1 (en) | Vision inspection method based on learning data | |
| CN117495891B (en) | Point cloud edge detection method and device and electronic equipment | |
| CN112317957B (en) | Laser welding method, laser welding apparatus, and storage medium therefor | |
| WO2021103875A1 (en) | Laser sensor-based plane detection method and device | |
| CN114299040A (en) | Ceramic tile flaw detection method and device and electronic equipment | |
| CN116091485A (en) | Flotation process foam image feature selection method and device based on sensitive mutual information | |
| CN110109945B (en) | AOI detection method and device applied to substrate, storage medium and AOI detection equipment | |
| CN116665009A (en) | Pipeline magnetic flux leakage image detection method based on multi-scale SSD network | |
| CN111062406A (en) | Semi-supervised optimal transmission method for heterogeneous field adaptation | |
| CN107067034B (en) | Method and system for rapidly identifying infrared spectrum data classification | |
| CN115618506A (en) | Method for predicting power of single-shaft combined cycle gas turbine | |
| CN115713484A (en) | Industrial defect sample detection method and system based on image representation | |
| CN104317912B (en) | Image meaning automatic marking method based on neighborhood and learning distance metric | |
| CN114037941A (en) | Method and device for performing algorithm multi-data cross validation completion aiming at video target attributes | |
| CN113191439A (en) | Deviation punishment enhanced stacking automatic encoder processing method and device | |
| CN117197224B (en) | Raman spectrometer self-adaptive focusing device and method based on residual error network | |
| CN116955737B (en) | Abnormal characteristic retrieval method used in gelatin production process |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20210205 Assignee: ZHONGSHAN AISCENT TECHNOLOGIES Co.,Ltd. Assignor: WUYI University Contract record no.: X2023980052830 Denomination of invention: Laser welding methods, devices, and storage media Granted publication date: 20221108 License type: Common License Record date: 20231219 |
|
| EE01 | Entry into force of recordation of patent licensing contract |